Data exports

Everything behind this site, free to download and reuse under CC BY 4.0, or embed a card for one plant or country. The fuel chain tables in particular do not exist in this form anywhere else. They were assembled by hand from agency reports and company filings, and every capacity figure carries the year it applies to and a link to its source.

File	Rows	Contents
`reactors.csv`	1,825	Every unit: status, type, fuel family, vendor, capacities, milestone dates, coordinates, and the date GEM last researched it. Design net capacity, model line, the model line's vendor and start year are the last four columns.
`sites.csv`	592	One row per site: position, operating and planned units, capacity, and, where the position is exact, elevation and distance to the sea or a large lake.
`capacity-by-year.csv`	-	Operating capacity at 31 December of each year, for the world and each country. Units with no start year and mothballed units are left out, so it is a floor.
`facilities.csv`	158	Mines, conversion, enrichment, fabrication, HALEU, TRISO, reprocessing and disposal, with capacity, the year each figure applies to, and its source.
`companies.csv`	84	Reactor vendors and fuel chain operators, with ownership type, listing, parent, and the roles each plays.
`smr-designs.csv`	26	Advanced reactor designs with regulator, docket, application type, review phase and credibility tier.
`supplies.csv`	21	Which fabricators can supply which fuel family, sourced.
`fuel-chain.json`	-	The whole fuel chain in one document: stages, world capacity by operator, facilities, companies and supply relationships.
`ownership.csv`	-	Every owner-reactor stake used to attribute operating capacity on /companies/ownership/, including the ones left out as not stated.
`fuel-exposure.csv`	-	Operating units per country by fuel family, documented fabricators, and exposure to Russia-only or undocumented supply.
`sites.json`	-	One exact coordinate per plant site, with country and unit counts. Powers the nearby-plants tool; straight-line distance only.
`corrections.csv`	54	Every correction applied on top of the source data: the field, its old and new value, why, and what was checked against.

Attribution

If you use this data, please attribute it. A link is enough:

NuclearChain (nuclearchain.net), CC BY 4.0.
Fleet data derived from Global Nuclear Power Tracker,
Global Energy Monitor, August 2026 release, CC BY 4.0.

What you should know before using it

Fleet status can lag. 93% of units the base dataset lists as operating were last researched in May 2024. Each row carries date_last_researched, and where we have corrected a record against a later source the change is documented on the page and in corrections.yaml in the repository.
Only gross capacity is complete. capacity_gross_mw is populated for every unit; net capacity is populated for about 40%. Summing net would silently understate. Do not.
Null means not disclosed. It never means zero, and it is never a placeholder for a value we guessed.
Mining rows carry output, not capacity. output_value is actual production for output_year; those rows have no nameplate capacity.
Some operators publish only group figures. Urenco reports one capacity across four plants and Rosatom across four. Those facility rows have no capacity rather than an invented per-site share.

Live status

US daily reactor power levels are served separately as JSON, refreshed daily from the NRC and cached for 15 minutes: /api/status.json. That feed is US Government work and in the public domain. It is one of several live feeds this site runs; see the full API reference for the rest, and data freshness for whether each one is currently up to date.